Bone response to hormone replacement therapy (HRT) was assessed in a retrospective longitudinal study. 115 women started on HRT for the first time and 252 controls underwent initial densitometric evaluation (spine and femur) and a second evaluation on average 29 months later. Patients were classified as post-menopausal or perimenopausal according to whether they had had more or less than 6 months amenorrhea. Oral or non-oral 17-beta estradiol was used at the dose generally accepted to be skeletally protective. In the post-menopausal group, prescription of HRT was followed by spinal and femoral bone gain (+ 2.85% and + 1.06% per year respectively). There was no bone gain with HRT in the peri-menopausal women, but the stability seen contrasted greatly with the very marked bone loss found in controls (spine - 3.09% per year and femur - 1.78% per year). Lumbar densitometric variations were correlated, at least in the post-menopausal group, with those in the femur, but the amplitude of femoral variations was half that of the spine. Body mass index (BMI) was not found to be a predictive factor of bone response to HRT in this group, but the time since the menopause and initial densitometric results were. For the spine, the % of subjects losing their bone mass in response to the start of HRT, nil in the post-menopausal group, was 16% in peri-menopausal women. The % of good responders increased from 8% peri-menopausally to more than 59% 2 years after the menopause. The response in the femur appeared to be very different, with 20% good responders and a % of stable subjects similar to that of the control group. Femoral variations and the existence regarding the spine of a group showing no or only a poor response to HRT would be in favor of densitometric monitoring when initial bone mass is low.